Who Owns Enterprise AI? The Governance Question Stalling Scale

Artificial intelligence has become an enterprise-wide capability, but a fundamental organizational question remains unresolved: Who actually owns it?

Unlike previous technology waves that naturally fell under IT's purview, AI touches nearly every business function. Contact centers, security teams, data organizations, line-of-business leaders, and IT departments all have legitimate claims to ownership. The result is governance models that fragment before AI programs reach meaningful scale.

How enterprises divide responsibility for AI strategy, governance, and execution will determine whether AI becomes a coordinated business capability or another layer of operational complexity.

Why it matters

This isn't an abstract org-chart debate. According to IDC research from June 2026, 68% of agentic AI investment decisions are now driven by business leaders, while just 4% are owned by AI centers of excellence. Many business units are building their own AI agents and introducing third-party tools outside traditional IT oversight. Without aligned governance, organizations risk financial surprises from usage-based pricing, compliance violations, security failures, and an inability to scale AI sustainably.

Business units are moving faster than IT

The ownership challenge stems from AI's fundamental accessibility. Unlike previous enterprise technology that required specialized expertise, AI enables anyone to use natural language to write, analyze, build, code, or automate work.

"Marketing, sales, finance, customer service, HR and operations are all adopting AI to solve business problems, blurring the traditional lines of ownership," said Juan Jaysingh, CEO of agentic workflow platform Zingtree. "The challenge isn't determining who owns AI — it's recognizing that everyone does."

Alessandro Perilli, vice president of enterprise AI strategies at IDC, points to a structural tension: business units must deliver financial results while the CIO office must govern risk. These divergent incentives create different speeds of adoption.

The growing independence of business units has led some organizations to adopt zero-trust approaches governing who can build and deploy AI.

Governance as enabler, not bottleneck

The most successful organizations are repositioning governance teams from gatekeepers to enablers, according to Jaysingh. This starts with expanding AI literacy across the workforce before allowing broader adoption, then methodically provisioning systems and data access so teams can safely deploy AI solutions.

Perilli notes there's no single organizational model emerging as standard because every enterprise balances innovation and risk differently. Organizations prioritizing speed may give business units responsibility for AI budgets and delivery while leaving governance to the CIO organization.

According to Perilli, 68% of organizations do not yet have a fully operational AI center of excellence, though those that do are moving significantly more AI pilots into production.

Organizations can reduce friction by establishing pre-approved budgets and low-risk AI use cases that require no additional review. Once business value has been validated, a standing automation team can polish and scale the initial implementation.

The cost visibility challenge

Governance now extends beyond compliance and security to include cost visibility. With enterprise AI shifting to usage-based pricing in 2026, organizations face potential financial surprises reminiscent of rising cloud bills from the last decade.

"Building that visibility and discipline now is what separates organizations that scale AI sustainably from the ones that get blindsided by a hefty bill, a compliance violation or a security failure," Jaysingh said.

Alignment matters more than ownership

Fragmented ownership isn't necessarily problematic if all stakeholders align on the same priorities, Perilli argues. The real danger emerges when alignment disappears, leading to slowed innovation, workforce alienation, loss of talent, uncontrolled risk, and inability to scale.

Still, Perilli cautions against assuming today's organizational structures will remain appropriate as AI continues to evolve. "It's risky to assume that organizational models that worked for the past 20 years will work in this new era," he said.

These details were first reported by Nathan Eddy for No Jitter.